The all-electron GW method based on WIEN2k:
Implementation and applications.
Ricardo I. G´omez-Abal
Fritz-Haber-Institut of the Max-Planck-Society
Faradayweg 4-6, D-14195, Berlin, Germany
15th. WIEN2k-Workshop
March, 29th. 2008
The all-electron GW method based on WIEN2k: Implementation and applications.
1. The all-electron GW method based on WIEN2k:
Implementation and applications.
Ricardo I. G´omez-Abal
Fritz-Haber-Institut of the Max-Planck-Society
Faradayweg 4-6, D-14195, Berlin, Germany
15th. WIEN2k-Workshop
March, 29th. 2008
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 1 / 54
3. Introduction
Density Functional Theory
E ⇐⇒ ρ
Kohn-Sham scheme
interacting electrons fictious particles
non interacting
Condition:
n(r) =
occ
i
|Ψi (r)|2
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 3 / 54
4. Introduction
Density Functional Theory
Kohn-Sham equation
[T + Vext + VH + Vxc]Ψi = ǫi Ψi
Vxc
local
energy independent
hermitian
ǫi :
Lagrange multipliers
No physical meaning
Exception: Highest occupied ǫi = −I
Fast.
Good structural properties.
Excitation spectra??
E ⇐⇒ ρR. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 4 / 54
5. Introduction
The Bandgap Problem
LDA vs. experimental bandgaps
Si
GaAs
zGaN
ZnS
C
CaO
MgO
NaCl
0 2 4 6 8
Eg
exp
[eV]
0
2
4
6
8Eg
LDA
[eV]
Up to 50% underestimation
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 5 / 54
6. Introduction
Many-Body Theory
Quasiparticle equation
[T + Vext + VH]Ψi (r) + Σ(r, r′
; ǫi )Ψi (r′
)d3
r′
= ǫi Ψi (r)
Σ
non local
energy dependent
non hermitian
ǫi :
Poles of the Green’s Function
ǫi =
E(N) − E(N − 1, i) ǫi < EF
E(N + 1, i) − E(N) ǫi > EF
Formally correspond to the excitation spectra
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 6 / 54
7. Introduction
Many Body Theory (cont.)
The Self-Energy (Σ)
Hedin, 1965: Expansion in terms of the dynamically screened Coulomb
potential (W ).
Fast convergence.
First order:
Σ(r, t, r′
, t′
) = G(r, t, r′
, t′
)W (r, t, r′
, t′
)
Simplest approximation including dynamical correlation effects
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 7 / 54
8. Introduction
Many Body Theory (cont.)
The Screened Coulomb potential (W )
W (r1, r2; ω) = ε−1
(r1, r3; ω)v(r3, r2)dr3
ε(r1, r2; ω) =1 − v(r1, r3)P(r3, r2; ω)dr3
P(r1, r2; ω) = −
i
2π
G(r1, r2; ω + ω′
)G(r2, r1; ω‘)dω′
Requires selfconsistency with the QP equation!
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 8 / 54
16. Implementation Basis Functions
The Polarization
P(r1, r2, ω) =
X
n,m,k,k′
Ψnk(r1)Ψ∗
mk′ (r1)Ψ∗
nk(r2)Ψmk′ (r2)F(ǫnk, ǫmk′ ; ω)
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 12 / 54
17. Implementation Basis Functions
The basis functions
P(r1, r2, ω) =
X
n,m,k,k′
Ψnk(r1)Ψ∗
mk′ (r1)Ψ∗
nk(r2)Ψmk′ (r2)F(ǫnk, ǫmk′ ; ω)
Definition
χq
i (r) =
Ra
eik·(R+ra)
υNL(r)YLM (ˆr) r ∈ MT-spheres
1
√
V G
Si,Gei(q+G)·r
r ∈ Interstitial
1 F. Aryasetiawan and O. Gunnarsson, Rep. Prog. Phys. 61, 237 (1998).
2 T. Kotani and M. van Schilfgaarde, Sol. State. Comm. 121, 461 (2002).
MT
Sphere
Interstitial
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 12 / 54
18. Implementation Basis Functions
The basis functions
Obtaining the radial functions
For each L take ul (r)ul′ (r) such that |l − l′| ≤ L ≤ l + l′
Calculate the overlap matrix:
Oa
ll′,l1l′
1
=
RMT
0
ul (r)ul′ (r)ul1
(r)ul′
1
(r)r2
dr
Solve the secular equation:
Oa
ll′,l1l′
1
− λnδll1
δl′l′
1
cn,l1l′
1
= 0
If λn ≥ λtol then:
υnL(r) =
ll′
cn,ll′ ul (r)ul′ (r)
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 13 / 54
19. Implementation Basis Functions
The basis functions
The matrix elements
Mi
nm(k, q) ≡
Ω
˜χq
i (r)Ψm,k−q(r)
∗
Ψnk(r)d3
r
Tests
Visual
Ψn,k(r)Ψ∗
m,k−q(r) =
i
Mi
nm(k, q)˜χq
i (r)
Completeness
∆ =
R
|Ψn,k(r)Ψ∗
m,k−q(r)−
P
i Mi
nm(k,q)˜χq
i (r)|2d3r
R
|Ψn,k(r)Ψ∗
m,k−q(r)|2d3r
∆ = 1 − i |Mi
nm(k, q)|2
|Ψn,k(r)Ψ∗
m,k−q(r)|2d3r
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 14 / 54
20. Implementation Basis Functions
Basis Functions: Tests
k = Γ, k′
= X, n = 4 m = 5
At1
RMT1
RMT2
At2
r
-0.04
-0.02
0
0.02
Ψnk
Ψmk’
Exact
Fit
Ψn,k(r)Ψ∗
m,k−q(r) =
X
i
Mi
nm(k, q)˜χq
i (r)
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 15 / 54
21. Implementation Basis Functions
Basis Functions: Tests
k = Γ, k′
= π
a (1, 0, 0), n = 2 m = 6
At1
RMT1
RMT2
At2
r
-0.01
0
0.01
Ψnk
Ψmk’
Exact
Fit
Ψn,k(r)Ψ∗
m,k−q(r) =
X
i
Mi
nm(k, q)˜χq
i (r)
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 15 / 54
22. Implementation Basis Functions
Basis Functions: Tests
k = 3π
2a (1, 1, −1), k′
= π
2a (1, −1, 1), n = 2, m =
6
At1
RMT1
RMT2
At2
r
-0.1
-0.05
0
0.05
0.1
Ψnk
Ψmk’
Exact
Fit
Ψn,k(r)Ψ∗
m,k−q(r) =
X
i
Mi
nm(k, q)˜χq
i (r)
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 15 / 54
23. Implementation Basis Functions
Basis Functions: Tests
k = Γ, k′
= π
a (1, 0, 0), n = 1, m = 5
At1
RMT1
RMT2
At2
r
0
0.5
1Ψnk
Ψmk’
Exact
Fit
Ψn,k(r)Ψ∗
m,k−q(r) =
X
i
Mi
nm(k, q)˜χq
i (r)
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 15 / 54
24. Implementation Basis Functions
Basis Functions: Completeness test.
MT-Spheres
20 40 60 80 100 120 140 160 180
Number of spherical basis functions per atom
1e-06
1e-05
1e-04
0.001
0.01
0.1
Relativeerror
tol.=1e-2
tol.=1e-3
tol.=1e-4
tol.=1e-5
tol.=1e-7
tol.=1e-2
tol.=1e-3
tol.=1e-4
tol.=1e-5
lmax=2 lmax=3lmax=1
tol.=1e-2
tol.=1e-3
tol.=1e-5
Minimum
Maximum
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 16 / 54
25. Implementation Basis Functions
Basis Functions: Completeness test.
MT-Spheres
20 40 60 80 100 120 140 160 180
Number of spherical basis functions per atom
1e-06
1e-05
1e-04
0.001
0.01
0.1
Relativeerror
tol.=1e-2
tol.=1e-3
tol.=1e-4
tol.=1e-5
tol.=1e-7
tol.=1e-2
tol.=1e-3
tol.=1e-4
tol.=1e-5
lmax=2 lmax=3lmax=1
tol.=1e-2
tol.=1e-3
tol.=1e-5
Minimum
Maximum
Remarks
Decreases with lmax .
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 16 / 54
26. Implementation Basis Functions
Basis Functions: Completeness test.
MT-Spheres
20 40 60 80 100 120 140 160 180
Number of spherical basis functions per atom
1e-06
1e-05
1e-04
0.001
0.01
0.1
Relativeerror
tol.=1e-2
tol.=1e-3
tol.=1e-4
tol.=1e-5
tol.=1e-7
tol.=1e-2
tol.=1e-3
tol.=1e-4
tol.=1e-5
lmax=2 lmax=3lmax=1
tol.=1e-2
tol.=1e-3
tol.=1e-5
Minimum
Maximum
Remarks
Decreases with lmax.
Saturates with ǫtol
for a given lmax .
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 16 / 54
27. Implementation Basis Functions
Basis Functions: Completeness test.
Interstitial
150 200 250 300 350
Number of interstitial basis functions
1e-08
1e-07
1e-06
1e-05
1e-04
0.001
Relativeerror
LAPWbasis
Minimum
Maximum
Remarks
Decreases with lmax.
Saturates with ǫtol for
a given lmax .
Decreases with
|Gmax|.
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 16 / 54
28. Implementation Basis Functions
Basis Functions: Completeness test.
Interstitial
150 200 250 300 350
Number of interstitial basis functions
1e-08
1e-07
1e-06
1e-05
1e-04
0.001
Relativeerror
LAPWbasis
Minimum
Maximum
Remarks
Decreases with lmax.
Saturates with ǫtol for
a given lmax .
Decreases with |Gmax|.
Recipe
Choose max(ǫtol ) that
saturates
Choose |Gmax| so that
∆i ≈ ∆MT
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 16 / 54
29. Implementation Brillouin Zone Integration
The Polarization
Pij (q, ω) =
Z
BZ
X
nm
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫmk−q+ǫnk+iη
−
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫnk+ǫmk−q−iη
ff
f (ǫk)[1 − f (ǫk−q)]d3
k
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 17 / 54
30. Implementation Brillouin Zone Integration
Brillouin Zone Integration
Pij (q, ω) =
Z
BZ
X
nm
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫmk−q+ǫnk+iη
−
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫnk+ǫmk−q−iη
ff
f (ǫk)[1 − f (ǫk−q)]d3
k
q-dependent Linear Tetrahedron Method
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 17 / 54
31. Implementation Brillouin Zone Integration
Brillouin Zone Integration
Pij (q, ω) =
Z
BZ
X
nm
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫmk−q+ǫnk+iη
−
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫnk+ǫmk−q−iη
ff
f (ǫk)[1 − f (ǫk−q)]d3
k
q-dependent Linear Tetrahedron Method
Linear Tetrahedron Method
P(q, ω) =
Z
BZ
X(k, q)
f (ǫk)[1 − f (ǫk−q)]
ω − ∆ǫ
d3
k
wT
ki ,q =
Z
ΩT
wi (k)
f (ǫk)[1 − f (ǫk−q)]
ω − ∆ǫ
d3
k wki ,q =
X
T∋ki
wT
ki ,q
P(q, ω) =
X
nm
X
i
X(ki , q)wnm(ki , q; ω)
J. Rath and A. J. Freeman, Phys. Rev. B 11, 2109 (1975)
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 17 / 54
32. Implementation Brillouin Zone Integration
Brillouin Zone Integration
Pij (q, ω) =
Z
BZ
X
nm
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫmk−q+ǫnk+iη
−
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫnk+ǫmk−q−iη
ff
f (ǫk)[1 − f (ǫk−q)]d3
k
q-dependent Linear Tetrahedron Method
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 18 / 54
33. Implementation Brillouin Zone Integration
Brillouin Zone Integration
Pij (q, ω) =
Z
BZ
X
nm
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫmk−q+ǫnk+iη
−
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫnk+ǫmk−q−iη
ff
f (ǫk)[1 − f (ǫk−q)]d3
k
q-dependent Linear Tetrahedron Method
q
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 18 / 54
34. Implementation Brillouin Zone Integration
Brillouin Zone Integration
Pij (q, ω) =
Z
BZ
X
nm
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫmk−q+ǫnk+iη
−
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫnk+ǫmk−q−iη
ff
f (ǫk)[1 − f (ǫk−q)]d3
k
q-dependent Linear Tetrahedron Method
q
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 18 / 54
35. Implementation Brillouin Zone Integration
Brillouin Zone Integration
Pij (q, ω) =
Z
BZ
X
nm
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫmk−q+ǫnk+iη
−
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫnk+ǫmk−q−iη
ff
f (ǫk)[1 − f (ǫk−q)]d3
k
q-dependent Linear Tetrahedron Method
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 19 / 54
36. Implementation Brillouin Zone Integration
Brillouin Zone integration
Pij (q, ω) =
Z
BZ
X
nm
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫmk−q+ǫnk+iη
−
Mi
nm(k,q)[Mj
nm(k,q)]∗
ω−ǫnk+ǫmk−q−iη
ff
f (ǫk)[1 − f (ǫk−q)]d3
k
q-dependent Linear Tetrahedron Method
4 nodes 6 nodes 8 nodes 8 nodes
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 20 / 54
37. Implementation Brillouin Zone Integration
q-dependent Linear Tetrahedron Method
Test: Free electron gas
0 0.5 1 1.5 2 2.5 3 3.5
q/kF
0
0.2
0.4
0.6
0.8
1
Lindhardtfunction
exact
364 k-pts
540 k-pts
1368 k-pts
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 21 / 54
38. Implementation Brillouin Zone Integration
q-dependent Linear Tetrahedron Method
Test: Cu bandstructure
W L Γ X Z W K
-10
EF
10 LDA
GW
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 22 / 54
39. Implementation Brillouin Zone Integration
q-dependent Linear Tetrahedron Method
Test: Cu DOS
-8 -6 -4 -2 EF
2 4
Energy [eV]
0
1
2
3
4
5
6DOS
LDA
GW
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 23 / 54
40. Implementation Brillouin Zone Integration
q-dependent Linear Tetrahedron Method
Test: Al bandstructure
W L Γ X Z W K
-10
EF
10
LDA
GW
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 24 / 54
41. Implementation The Γ-point singularity
The Γ point singularity
The symmetrized dielectric matrix
Definition
˜εij (q, ω) =
lm
v
1
2
il (q)Plm(q, ω)v
1
2
mj (q)
The screened potential
Wij (q, ω) =
lm
v
1
2
il (q)˜ε−1
lm (q, ω)v
1
2
mj (q)
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 25 / 54
42. Implementation The Γ-point singularity
The Γ point singularity
The screened Coulomb potential
v
1
2
ij (q → 0) =
v
s 1
2
ij
|q|
+ ˜v
1
2
ij (q)
Wij (q, ω) =
1
|q|2
W s2
ij (q, ω) +
1
|q|
W s1
ij (q, ω) + ˜Wij (q, ω)
Diverges!
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 26 / 54
43. Implementation The Γ-point singularity
The Γ point singularity
The screened Coulomb potential
v
1
2
ij (q → 0) =
v
s 1
2
ij
|q|
+ ˜v
1
2
ij (q)
Wij (q, ω) =
1
|q|2
W s2
ij (q, ω) +
1
|q|
W s1
ij (q, ω) + ˜Wij (q, ω)
But can be integrated
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 26 / 54
44. Implementation Frequency convolution
Implementation
Frequency convolution
Σ(r1, r2; ω) =
i
2π
Z
G0(r1, r2; ω + ω′
)W0(r2, r1; ω‘)dω′
⇓ Analytic continuation
Σ(r1, r2; iω) =
i
2π
Z
G0(r1, r2; iω + iω′
)W0(r2, r1; iω‘)diω′
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 27 / 54
45. Implementation Frequency convolution
Implementation
Frequency convolution
Σ(r1, r2; ω) =
i
2π
Z
G0(r1, r2; ω + ω′
)W0(r2, r1; ω‘)dω′
⇓ Analytic continuation
Σ(r1, r2; iω) =
i
2π
Z
G0(r1, r2; iω + iω′
)W0(r2, r1; iω‘)diω′
Σnk(iω) = −
X
q
X
ij
X
n′
Mi
nn′ (k, q)
1
π
∞Z
0
Wij (q; iω′)dω′
(iω − ǫn,k′ )2 + ω′2
M∗j
n′n
(k, q)
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 27 / 54
46. Implementation Frequency convolution
Implementation
Frequency convolution
Σ(r1, r2; ω) =
i
2π
Z
G0(r1, r2; ω + ω′
)W0(r2, r1; ω‘)dω′
⇓ Analytic continuation
Σ(r1, r2; iω) =
i
2π
Z
G0(r1, r2; iω + iω′
)W0(r2, r1; iω‘)diω′
Σnk(iω) = −
X
q
X
ij
X
n′
Mi
nn′ (k, q)
1
π
∞Z
0
Wij (q; iω′)dω′
(iω − ǫn,k′ )2 + ω′2
M∗j
n′n
(k, q)
Pad´e Approximant
Σnk(iω) =
PN
j=0 aj (iω)j
PN+1
j=0 bj (iω)j
⇓ Analytic continuation
Σnk(ω) =
PN
j=0 aj ωj
PN+1
j=0 bj ωj
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 27 / 54
47. Implementation Summary
G0W0@Wien2k: A FP-(L)APW+lo + G0W0 code
Flowchart
Wien2k Begin
ψkn χiq
ǫDFT
kn Mi
nm(k, q) vij (q)
Pij (q, ω)
εij (q, ω)
Wij (q, ω)
V xc
k,n Σnn(k, ω)
ǫ
qp
k,n
End
Code keywords
Based on FP-(L)APW+lo
Mixed Basis
Linear Tetrahedron
Method
Reciprocal space
Imaginary frequencies
Excelent results
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 28 / 54
48. Implementation Convergence Tests
Silicon: Convergence tests
Number of frequencies
10 15 20 25
Nr. of frequencies
0.878
0.879
0.88
0.881
0.882
0.883
0.884
Eg
[eV]
Used paremeters
16 frequencies
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 29 / 54
49. Implementation Convergence Tests
Silicon: Convergence tests
Number of excited states
8 27 64 125
Nr. of k-points
0.45
0.46
0.47
0.48
0.49
0.5
0.51
0.52
0.53
∆Eg
[eV]
Used paremeters
16 frequencies
64 k-points
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 29 / 54
50. Implementation Convergence Tests
Silicon: Convergence tests
Number of k-points
0 50 100 150 200 250
Number of unoccupied bands
0.95
1
1.05
1.1
BandGap
Used paremeters
16 frequencies
64 k-points
∼ 200 unocc. bands
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 29 / 54
51. Results Bandgaps
Results I:
Bandgaps Si
GaAs
zGaN
ZnS
C
MgO
NaCl
CaO
0 2 4 6 8
Experimental Eg
[eV]
0
2
4
6
8
10
CalculatedEg
[eV]
G0
W0
All electron
G0
W0
Pseudopotentials
G0
W0
@Wien2k
LDA
- F. Aryasetiawan and O.
Gunnarson, Rep. Prog. Phys.
61, 237 (1998).(and Refs.)
- T. Kotani and M. van
Schilfgaarde, Solid State Comm.
121, 461 (2002).
- C. Friedrich et al, Phys. Rev. B
74, 045104 (2006).
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 30 / 54
52. Results Bandstructures
Results II:
Band diagrams: Silicon
W ΓL Λ ∆ X Z W K
-10
-5
εF
5
10
1
LDA
GW
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 31 / 54
55. Core-valence interaction Motivation
Debate
G0W0 Si bandgap
1990 2000 2010
Publication year
0
0.5
1
1.5
Eg
[eV]
Pseudopotentials
All electron
?
- M. Tiago et al, Phys. Rev. B 69, 125212 (2004).
- C. Friedrich et al, Phys. Rev. B 74, 045104 (2006).
Pseudopotentials
systematically larger
gaps
in better agreement
with expeeriment
All electron
Benchmark for ab-initio
calculations
- A. Fleszar, Phys. Rev. B 64, 245204
(2001).
- T. Kotani and M. van Schilfgaarde,
Solid State Comm. 121, 461 (2002).
- W. Ku and A. Eguiluz, Phys. Rev. Lett.
89,126401 (2002).
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 33 / 54
56. Core-valence interaction Core-valence Partitioning
Core-valence partitioning
Kohn-Sham equation
HKS Ψnk = ǫnkΨnk
All electron
HKS = T + Vnuc + Vh + Vxc[n]
Pseudopotentials
HKS = T + Vps + Vh + Vxc[˜nv + ˜nc]
Vxc[n] = Vxc[˜nv + ˜nc] + Vxc[ncore − ˜nc] ⊂ Veff
G0W0 correction
ǫqp
nk = ǫKS
nk + ∆ǫnk
All electron
∆ǫnk = ℜ (Σnk [{Ψnk; Ψc}]) − V xc
nk [n]
Pseudopotentials
∆ǫnk = ℜ(Σnk[{˜Ψnk}])−V xc
nk [˜nv +˜nc]
ℜ (Σnk [{Ψnk; Ψc}]) = ℜ(Σnk[{˜Ψnk}]) + V xc
nk [ncore − ˜nc] ⊂ Veff
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 34 / 54
57. Core-valence interaction Different Approaches
Different Approaches
G0W0 correction
All electron:
ǫqp
nk = ǫKS
nk + ℜ (Σnk [{Ψnk; Ψcore}]) − V xc
nk [n]
Pseudopotentials:
ǫqp
nk = ǫKS
nk + ℜ(Σnk[{˜Ψnk}]) − V xc
nk [˜nval ]
Valence only:
ǫqp
nk = ǫKS
nk + ℜ (Σnk [{Ψnk}]) − V xc
nk [nval ]
Separated terms
Σnk = Σx
nk + Σc
nk
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 35 / 54
63. Core-valence interaction GaAs
Results
GaAs: Matrix elements
Remarks
ΣC : PP ≈ Val ≈ AE
ΣX : PP ≈ Val = AE
VXC : PP ≈ Val = AE
Core-valence lin.
ΣC : Small effect
ΣX : Large effect
VXC : Large effect
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 40 / 54
64. Core-valence interaction GaAs
Results
GaAs: Matrix elements
Remarks
ΣC : PP ≈ Val ≈ AE
ΣX : PP ≈ Val = AE
VXC : PP ≈ Val = AE
ΣX − VXC : PP ≈ Val =
AE
Core-valence lin.
ΣC : Small effect
ΣX : Large effect
VXC : Large effect
ΣX − VXC :Large effect
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 40 / 54
65. Core-valence interaction GaAs
Results
GaAs gap: G0W0 correction
Remarks
∆Eg : PP < Val < AE
Core-valence lin.
∆Eg : Large effect.
Reduces the correction!!
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 41 / 54
66. Core-valence interaction GaAs
Conclusions
Core-valence partitioning:
Strong changes in ΣX and VXC
Small changes in ΣC
∆Eg : Small changes in Si. Large in GaAs
Does NOT systematically increase the G0W0-correction.
“Pseudoization” also plays an important role.
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 42 / 54
67. Core-valence interaction Semicore States
Semicore states
Semicore states binding energies
MgO 6p-5p1s transition energy
TotalW L Λ Γ ∆ X ZW K
-50
-40
-30
-20
-10
EF
10
Energy[eV]
sMg
pMg
s0
pO
LDA:=43.55eV
GW:=52.41eV
Excitation energy [eV]
ROHF(∆SCF)1
54.6
CASTP21
53.8
LDA2
43.5
G0W 2
0 52.4
Experiment3
53.4
1.- C. Sousa et al. Phys. Rev. B 62,
10013 (2000).
2.- This work.
3.- W. L. OBrien et al. Phys. Rev. B
44, 1013 (1991).
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 43 / 54
68. f-electron systems
f-electron systems
Motivation
Two strongly interrelated subsystems
itinerant spd states
strongly localized f states
Intriguing physics
Heavy fermion
Kondo effect
Etc...
Challenge to first-principle calculations:
LDA/GGA good for itinerant electrons
Fails for f -electrons
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 44 / 54
69. f-electron systems
(no)f-electron systems
CeO2
-6 -4 -2 0 2 4 6 8 10 12
Energy [eV]
0
5
10
15
DOS LDA
LDA-G0
W0
XPS+BIS
XPS+XAS
E. Wuilloud et al. Phys. Rev. Lett 53, 202 (1984)
D. R. Mullins et al. Surf. Sci. 409, 307 (1998)
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 45 / 54
75. Conclusions
Ongoing work: Spin polarized metals
Test: Ni
W L Γ X ZW K
-10
-5
EF
5
Energy[eV]
W L Γ X ZW K
UP DOWN
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 49 / 54
79. Conclusions
Last but not least
Future plans
Anisotropy of εmacro
Half metals
Efficiency Improvement
COHSEX@Wien2k + GW
BSE
QPscGW
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 52 / 54
80. Conclusions
Acknowledgements
Xinzheng Li (FHI, Berlin): Code development, LTM library.
Dr. Hong Jiang (FHI, Berlin): Code improvement, Spin polarization,
LDA+U.
Prof. Claudia Ambrosch-Draxl (MUL; Austria): Wien2k interface and
more...
Christian Meisenbichler (MUL; Austria): MPI Paralelization
Patrick Rinke and Christoph Freysoldt (FHI, Berlin):
Pseudopotentials, etc..
The boss
Matthias Scheffler
R. G´omez-Abal (FHI-Berlin) GW@Wien2k Vienna 2008 53 / 54